I run a cluster. I buildpkg on one and pkgonly on all the nodes. libperl refuses to use the package and will configure/make/check/install/blabla... This breaks stuffs when I start the merge from pkgonly on all the other nodes that will eventualy colide on trying to untar a libperl while it is rebuilded by another merge on another node. Reproducible: Always Steps to Reproduce: 1. 2. 3.
Can you provide example output and ls packages/*perl* ?
I won't provide an example output, but you can easily verify this yourself. Enable the buildpkg in FEATURES, emerge once to create the tarball package, re-emerging with the -k switch should use that tarball package, but it does not. Maybe it's Portage, I don't know. All I know is that I have to emerge libperl sequentialy on all the nodes, because the tarball is recreated each time, and eventualy a node will untar whiel another is taring and BOOM!.
# emerge -kvp libperl | grep binary [binary R ] sys-devel/libperl-5.8.6 +berkdb -debug -gdbm -ithreads (-uclibc) Please provide the output as Nick asked...
I have just encounter the same problem with binutils. So maybe it's more a limitation with Make, distcc or the like. With MAKEOPTS=j32, biutils failed. I then lowered it to -j24 and specified the IPs in /etc/distcc/hosts one by one treee time each (instead of IP/3), but still failed. It worked when I did MAKEOPTS-j3 emerge binutils. So I guess I could recompile again and again, incrementing MAKEOPTS by 1 each time to find the upper limit. My point is that there seem to be such a limit, but I don't know what modulates that, because most of the packages builds fine and parallelize well.
Oups! Man :( Wrong bug. I'm sorry. This has nothing to do with the bug mentionned about libperl, which is that emerge -k libperl wont use the buildpkg version and instead compile from scratch. I'm so sorry :(
